Transcript Diapositiva 1

Data Mining
VS
Visualization
Santiago González Tortosa
<[email protected]>
Contents
I. Data Mining VS Visualization
II. Visualize to DM
III. DM to Visualize (to DM)
IV. Real world work:
I.
Global Behavior Modeling: A New approach to Grid
autonomic management
2
Data Mining VS Visualization
• Data Mining
– Knowledge discovery and extration
– Not always is easy to see patterns, distributions, etc.
• Visualization
– Represents data (2D, 3D, Virtual Reality,…)
– Helps to extract patterns
– Not always is easy to represent data in 2 or 3
dimensions
3
Visualize to DM
• Visualization help us to extract any pattern in the
data
4
Visualize to DM
• Visualization help us to extract any pattern in the
data
5
DM to Visualize
• Data contains N (> 3) features
– Curse of Dimensionality
• We want to visualize all data
• Dimensionality Reduction
– Reduce number of features
– Transform and create new features
6
DM to Visualize
• Dimensionality Reduction
– L.J.P. van der Maaten, E.O. Postma, and H.J. van den
Herik. Dimensionality Reduction: A Comparative Review.
Tilburg University Technical Report, TiCC-TR 2009-005,
2009
• Convex techniques: optimize an objective function
that does not contain any local optima
• Nonconvex techniques: optimize objective functions
that do contain local optima
7
DM to Visualize
• Optimization techniques (hill climbing, evolutive,
etc.)
8
DM to Visualize
• Optimization techniques
– One objective
– One objective with constraints (Semi-Supervised and
labeling)
– Multiobjective
9
DM to Visualize
• Example: Optimize axis
10
DM to Visualize
• Dimensionality Reduction in 2 phases:
– FSS: Feature Subset Selection (wrapper, needed CLASS!)
– Transformation and creation of new features (f.e. PCA)
11
DM to Visualize
• Example of Dimensionality Reduction in 2 phases
– User expert interacts
12
DM to Visualize
• DM to Visualize….to DM!!
• The idea is to obtain new knowledge or patterns
viewing the data.
– Supervised info: data with the same class are represented
in the same area (KNN).
– Unsupervised info: data is agrouped
13
DM to Visualize
• Example that some data is agrouped
14
DM to Visualize
• Visualization
– 2D and 3D visualization
– Virtual Reality
• Inmersion
• Interaction
• Imagination
– Augmented Reality
15
Real world work
Global Behavior Modeling: A New
approach to Grid autonomic
management
Jesus Montes <[email protected]>
16
Data Mining
VS
Visualization
Santiago González Tortosa
<[email protected]>